2 * Copyright 2013 Facebook, Inc.
4 * Licensed under the Apache License, Version 2.0 (the "License");
5 * you may not use this file except in compliance with the License.
6 * You may obtain a copy of the License at
8 * http://www.apache.org/licenses/LICENSE-2.0
10 * Unless required by applicable law or agreed to in writing, software
11 * distributed under the License is distributed on an "AS IS" BASIS,
12 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
13 * See the License for the specific language governing permissions and
14 * limitations under the License.
17 #ifndef FOLLY_BASE_STRING_H_
18 #define FOLLY_BASE_STRING_H_
22 #include <boost/type_traits.hpp>
25 # include <ext/hash_set>
26 # include <ext/hash_map>
29 #include "folly/Conv.h"
30 #include "folly/FBString.h"
31 #include "folly/FBVector.h"
32 #include "folly/Range.h"
33 #include "folly/ScopeGuard.h"
35 // Compatibility function, to make sure toStdString(s) can be called
36 // to convert a std::string or fbstring variable s into type std::string
37 // with very little overhead if s was already std::string
41 std::string toStdString(const folly::fbstring& s) {
42 return std::string(s.data(), s.size());
46 const std::string& toStdString(const std::string& s) {
50 // If called with a temporary, the compiler will select this overload instead
51 // of the above, so we don't return a (lvalue) reference to a temporary.
53 std::string&& toStdString(std::string&& s) {
58 * C-Escape a string, making it suitable for representation as a C string
59 * literal. Appends the result to the output string.
61 * Backslashes all occurrences of backslash and double-quote:
65 * Replaces all non-printable ASCII characters with backslash-octal
69 * Note that we use backslash-octal instead of backslash-hex because the octal
70 * representation is guaranteed to consume no more than 3 characters; "\3760"
71 * represents two characters, one with value 254, and one with value 48 ('0'),
72 * whereas "\xfe0" represents only one character (with value 4064, which leads
73 * to implementation-defined behavior).
75 template <class String>
76 void cEscape(StringPiece str, String& out);
79 * Similar to cEscape above, but returns the escaped string.
81 template <class String>
82 String cEscape(StringPiece str) {
89 * C-Unescape a string; the opposite of cEscape above. Appends the result
90 * to the output string.
92 * Recognizes the standard C escape sequences:
94 * \' \" \? \\ \a \b \f \n \r \t \v
98 * In strict mode (default), throws std::invalid_argument if it encounters
99 * an unrecognized escape sequence. In non-strict mode, it leaves
100 * the escape sequence unchanged.
102 template <class String>
103 void cUnescape(StringPiece str, String& out, bool strict = true);
106 * Similar to cUnescape above, but returns the escaped string.
108 template <class String>
109 String cUnescape(StringPiece str, bool strict = true) {
111 cUnescape(str, out, strict);
116 * URI-escape a string. Appends the result to the output string.
118 * Alphanumeric characters and other characters marked as "unreserved" in RFC
119 * 3986 ( -_.~ ) are left unchanged. In PATH mode, the forward slash (/) is
120 * also left unchanged. In QUERY mode, spaces are replaced by '+'. All other
121 * characters are percent-encoded.
123 enum class UriEscapeMode : unsigned char {
124 // The values are meaningful, see generate_escape_tables.py
129 template <class String>
130 void uriEscape(StringPiece str,
132 UriEscapeMode mode = UriEscapeMode::ALL);
135 * Similar to uriEscape above, but returns the escaped string.
137 template <class String>
138 String uriEscape(StringPiece str, UriEscapeMode mode = UriEscapeMode::ALL) {
140 uriEscape(str, out, mode);
145 * URI-unescape a string. Appends the result to the output string.
147 * In QUERY mode, '+' are replaced by space. %XX sequences are decoded if
148 * XX is a valid hex sequence, otherwise we throw invalid_argument.
150 template <class String>
151 void uriUnescape(StringPiece str,
153 UriEscapeMode mode = UriEscapeMode::ALL);
156 * Similar to uriUnescape above, but returns the unescaped string.
158 template <class String>
159 String uriUnescape(StringPiece str, UriEscapeMode mode = UriEscapeMode::ALL) {
161 uriUnescape(str, out, mode);
166 * stringPrintf is much like printf but deposits its result into a
167 * string. Two signatures are supported: the first simply returns the
168 * resulting string, and the second appends the produced characters to
169 * the specified string and returns a reference to it.
171 std::string stringPrintf(const char* format, ...)
172 __attribute__ ((format (printf, 1, 2)));
174 /** Similar to stringPrintf, with different signiture.
176 void stringPrintf(std::string* out, const char* fmt, ...)
177 __attribute__ ((format (printf, 2, 3)));
179 std::string& stringAppendf(std::string* output, const char* format, ...)
180 __attribute__ ((format (printf, 2, 3)));
183 * Backslashify a string, that is, replace non-printable characters
184 * with C-style (but NOT C compliant) "\xHH" encoding. If hex_style
185 * is false, then shorthand notations like "\0" will be used instead
186 * of "\x00" for the most common backslash cases.
188 * There are two forms, one returning the input string, and one
189 * creating output in the specified output string.
191 * This is mainly intended for printing to a terminal, so it is not
192 * particularly optimized.
194 * Do *not* use this in situations where you expect to be able to feed
195 * the string to a C or C++ compiler, as there are nuances with how C
196 * parses such strings that lead to failures. This is for display
197 * purposed only. If you want a string you can embed for use in C or
198 * C++, use cEscape instead. This function is for display purposes
201 template <class String1, class String2>
202 void backslashify(const String1& input, String2& output, bool hex_style=false);
204 template <class String>
205 String backslashify(const String& input, bool hex_style=false) {
207 backslashify(input, output, hex_style);
212 * Take a string and "humanify" it -- that is, make it look better.
213 * Since "better" is subjective, caveat emptor. The basic approach is
214 * to count the number of unprintable characters. If there are none,
215 * then the output is the input. If there are relatively few, or if
216 * there is a long "enough" prefix of printable characters, use
217 * backslashify. If it is mostly binary, then simply hex encode.
219 * This is an attempt to make a computer smart, and so likely is wrong
222 template <class String1, class String2>
223 void humanify(const String1& input, String2& output);
225 template <class String>
226 String humanify(const String& input) {
228 humanify(input, output);
233 * Same functionality as Python's binascii.hexlify. Returns true
234 * on successful conversion.
236 * If append_output is true, append data to the output rather than
239 template<class InputString, class OutputString>
240 bool hexlify(const InputString& input, OutputString& output,
244 * Same functionality as Python's binascii.unhexlify. Returns true
245 * on successful conversion.
247 template<class InputString, class OutputString>
248 bool unhexlify(const InputString& input, OutputString& output);
251 * A pretty-printer for numbers that appends suffixes of units of the
252 * given type. It prints 4 sig-figs of value with the most
255 * If `addSpace' is true, we put a space between the units suffix and
259 * PRETTY_TIME - s, ms, us, ns, etc.
260 * PRETTY_BYTES_METRIC - kB, MB, GB, etc (goes up by 10^3 = 1000 each time)
261 * PRETTY_BYTES - kB, MB, GB, etc (goes up by 2^10 = 1024 each time)
262 * PRETTY_BYTES_IEC - KiB, MiB, GiB, etc
263 * PRETTY_UNITS_METRIC - k, M, G, etc (goes up by 10^3 = 1000 each time)
264 * PRETTY_UNITS_BINARY - k, M, G, etc (goes up by 2^10 = 1024 each time)
265 * PRETTY_UNITS_BINARY_IEC - Ki, Mi, Gi, etc
267 * @author Mark Rabkin <mrabkin@fb.com>
274 PRETTY_BYTES = PRETTY_BYTES_BINARY,
275 PRETTY_BYTES_BINARY_IEC,
276 PRETTY_BYTES_IEC = PRETTY_BYTES_BINARY_IEC,
280 PRETTY_UNITS_BINARY_IEC,
285 std::string prettyPrint(double val, PrettyType, bool addSpace = true);
288 * Write a hex dump of size bytes starting at ptr to out.
290 * The hex dump is formatted as follows:
292 * for the string "abcdefghijklmnopqrstuvwxyz\x02"
293 00000000 61 62 63 64 65 66 67 68 69 6a 6b 6c 6d 6e 6f 70 |abcdefghijklmnop|
294 00000010 71 72 73 74 75 76 77 78 79 7a 02 |qrstuvwxyz. |
296 * that is, we write 16 bytes per line, both as hex bytes and as printable
297 * characters. Non-printable characters are replaced with '.'
298 * Lines are written to out one by one (one StringPiece at a time) without
301 template <class OutIt>
302 void hexDump(const void* ptr, size_t size, OutIt out);
305 * Return the hex dump of size bytes starting at ptr as a string.
307 std::string hexDump(const void* ptr, size_t size);
310 * Return a fbstring containing the description of the given errno value.
311 * Takes care not to overwrite the actual system errno, so calling
312 * errnoStr(errno) is valid.
314 fbstring errnoStr(int err);
317 * Return the demangled (prettyfied) version of a C++ type.
319 * This function tries to produce a human-readable type, but the type name will
320 * be returned unchanged in case of error or if demangling isn't supported on
323 * Use for debugging -- do not rely on demangle() returning anything useful.
325 * This function may allocate memory (and therefore throw).
327 fbstring demangle(const char* name);
328 inline fbstring demangle(const std::type_info& type) {
329 return demangle(type.name());
333 * Debug string for an exception: include type and what().
335 inline fbstring exceptionStr(const std::exception& e) {
336 return folly::to<fbstring>(demangle(typeid(e)), ": ", e.what());
339 inline fbstring exceptionStr(std::exception_ptr ep) {
341 std::rethrow_exception(ep);
342 } catch (const std::exception& e) {
343 return exceptionStr(e);
345 return "<unknown exception>";
350 * Split a string into a list of tokens by delimiter.
352 * The split interface here supports different output types, selected
353 * at compile time: StringPiece, fbstring, or std::string. If you are
354 * using a vector to hold the output, it detects the type based on
355 * what your vector contains. If the output vector is not empty, split
356 * will append to the end of the vector.
358 * You can also use splitTo() to write the output to an arbitrary
359 * OutputIterator (e.g. std::inserter() on a std::set<>), in which
360 * case you have to tell the function the type. (Rationale:
361 * OutputIterators don't have a value_type, so we can't detect the
362 * type in splitTo without being told.)
366 * std::vector<folly::StringPiece> v;
367 * folly::split(":", "asd:bsd", v);
369 * std::set<StringPiece> s;
370 * folly::splitTo<StringPiece>(":", "asd:bsd:asd:csd",
371 * std::inserter(s, s.begin()));
373 * Split also takes a flag (ignoreEmpty) that indicates whether adjacent
374 * delimiters should be treated as one single separator (ignoring empty tokens)
375 * or not (generating empty tokens).
378 template<class Delim, class String, class OutputType>
379 void split(const Delim& delimiter,
381 std::vector<OutputType>& out,
382 bool ignoreEmpty = false);
384 template<class Delim, class String, class OutputType>
385 void split(const Delim& delimiter,
387 folly::fbvector<OutputType>& out,
388 bool ignoreEmpty = false);
390 template<class OutputValueType, class Delim, class String,
391 class OutputIterator>
392 void splitTo(const Delim& delimiter,
395 bool ignoreEmpty = false);
398 * Join list of tokens.
400 * Stores a string representation of tokens in the same order with
401 * deliminer between each element.
404 template <class Delim, class Iterator, class String>
405 void join(const Delim& delimiter,
410 template <class Delim, class Container, class String>
411 void join(const Delim& delimiter,
412 const Container& container,
414 join(delimiter, container.begin(), container.end(), output);
417 template <class Delim, class Container>
418 std::string join(const Delim& delimiter,
419 const Container& container) {
421 join(delimiter, container.begin(), container.end(), output);
427 // Hash functions for string and fbstring usable with e.g. hash_map
429 namespace __gnu_cxx {
432 struct hash<folly::basic_fbstring<C> > : private hash<const C*> {
433 size_t operator()(const folly::basic_fbstring<C> & s) const {
434 return hash<const C*>::operator()(s.c_str());
439 struct hash<std::basic_string<C> > : private hash<const C*> {
440 size_t operator()(const std::basic_string<C> & s) const {
441 return hash<const C*>::operator()(s.c_str());
445 } // namespace __gnu_cxx
448 // Hook into boost's type traits
451 struct has_nothrow_constructor<folly::basic_fbstring<T> > : true_type {
452 enum { value = true };
456 #include "folly/String-inl.h"